upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-23 11:10:55 -05:00
Code Issues Packages Projects Releases Wiki Activity

chat : fix reasoning leak with force-opened bare <think> templates (#24674)

Browse Source 
* chat : fix reasoning leak with force-opened bare <think> templates

The reasoning start tag inferred from prior turns can carry trailing
whitespace (e.g. <think>\n) while a force-open template prefills a bare
<think>. Trim the tag used for the prefix split so the bare prefill is
matched instead of being swallowed into content.

* chat : fix Nemotron Nano v2 regression

---------

Co-authored-by: Alde Rojas <hello@alde.dev>
This commit is contained in:
Frosty40
2026-07-13 02:45:10 -05:00
committed by GitHub
parent efb3036c18
commit 91c631b21d
3 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
common/chat-auto-parser-generator.cpp
View File

@@ -147,7 +147,8 @@ common_peg_arena autoparser::build_parser(const generation_params & inputs, cons
} else {
parser = content.build_parser(ctx);
}
return pure_content ? p.prefix(generation_prompt, reasoning.start) + parser : p.prefix(generation_prompt, reasoning.start) << parser;
const std::string reasoning_start = trim_whitespace(reasoning.start);
return pure_content ? p.prefix(generation_prompt, reasoning_start) + parser : p.prefix(generation_prompt, reasoning_start) << parser;
});
}